Quote:

Originally Posted by **DionysusToast** Over time, my guess is this will blow up Howards account, although I hope this does not happen.
The trades are based around a mathematically derived probability of certain prices being hit. From what I understand so far, It appears that no market analysis is being performed, rather that if a TOS indicator says that the probability is below a certain threshold then the trade will be taken.
Remember that I have not studied the indicator in question but my presumption is that it will be such that it implies the probability of a losing streak is extremely low - perhaps 1 in a million. Whatever it does imply, it is just a formula and we have seen time and time again the way these formulas fail in real life.
When going over them in retrospect, there appear to be 2 common flaws, certainly in the models that have caused major crashes. The first flaw is in the assumptions fed into the model, these models need inputs and the way some of those are derived looks fairly dumb in retrospect. Obviously the inputs are based on what worked in the past. But assumptions like "house prices will rise 12% per annum indefinitely IS dumb". The second flaw is in failing to grasp the impact the use of the model has on the markets, a model that gets adopted industry wide is never adjusted for the fact that this new model is now going to impact the markets. I presume the TOS model is not used industry wide - so perhaps that issue is not going to be important here.
It appears this trading approach cannot sustain a losing streak. Howard mentions a 30% drawdown if he has a bad months. What about 3 bad months ? It may be statistically impossible but that hasn't stopped these things occuring in the past.
I hope I am wrong - both for Howard and his students but this to me sounds like a Black Swan waiting to happen.
If I am wrong, then Howard has found the Holy Grail. A purely mathematical approach to trading the markets with no technical or fundamental analysis required that will yield a consistent 10% a month.
Good luck. |

Appropriate concerns all. Like any business I've run, I have a process of quality control of myself and my strategy. Some of these I've mentioned earlier in this thread, but they likely bear repeating. However, lets discuss each of your points in order and see where we go from here.

**Probability of Touching**
I can't speak to the TOS proprietary model, but I can speak to the one I developed that came close enough to the TOS one to abandon mine and adapt theirs. My inputs were, current price of underlying instrument, strike price, interest rates, days to expiration and volatility. These are the same inputs that TOS reports that they use.

There are quite a few probability models used in options trading, all of which answer slightly different questions but all of which use the same inputs. I don't think the one I am using is any more magic or any less magic than the others.

These models are tools to be used carefully and with attention to the quality of the result they produce to see if the tool is still effective. If the quality assurance process is effective, any change in quality should be detectable before disaster occurs.

**Loosing streak**
From what I have learned from you, you are probably better at probability stuff than I. My understanding of these probability models is that they can be used as a proxy of the chance of the event happening. If I start with a probability of touching of 10%, then my understanding is that there is only a 10% chance, given the current market conditions, that the underlying instrument price will reach that level.

Were I to use a "set it and forget it" approach to trading, then it's pretty simple to calculate the probability of two successive failure events taking place. Since I actively manage the spreads, the probability of my loosing my maximum limit is less than is implied by the initial probability of touching would indicate. And so is the probability of two successive monthly losses.

Perhaps my explanation of how I derived my estimates was too convoluted because the 30% example was to represent only one half of the funds at risk. To reach the 30% level, would require two consecutive failure months. I tried to use this worst case (one that I never saw in testing) to provide a very conservative estimate of the expected return when losing months were included.

**Common flaws in models**
All models are based on assumptions and the one I use is not immune from that. All models must be monitored to see if they are effective under current market conditions. The model I use is not immune from that either. I fail to see how my model is more or any less susceptible to these challenges.

**Quality Assurance**
In my first post in this thread I described the reason for and the use of the journal. It is to monitor the two key components of a trading system, the strategy and the trader. There is a formal process for detecting degradations of performance of the trader separate from the model. These are designed to identify problems before they create a financial disaster.

**Conclusion**
Will this all continue to work as it does now? Will I be able to detect system failure before there is serious impact on my account? Have I thought of everything?

I don't know. I believe so. I doubt it. In that order.

One of the primary reasons that I started this thread was to learn from those with more experience and wisdom than I. The more I'm challenged, the deeper I have to think, the more I understand the strengths and weaknesses of my approach.

I greatly appreciate all the thoughtful questions that come with this thread.